Jun Chen - Research

My research interests focus on the design, analysis and implementation of control and optimization algorithms for large-scale autonomous systems. More specifically, I research novel decision-making algorithms and use these algorithms to build planning modules that enable complex systems to operate autonomously safely and efficiently. Motivated by the goal of building practical methodology for robust real-time decision-making, a research program is built for computational control and optimization, which could provide complex systems reliable, accurate, computational efficient and robust solutions.

Real-time Decision-making Algorithms and Tools for Autonomous Systems

Real-time decision-making is critical for the autonomous systems due to their dynamic nature. An efficient computational framework with the proper model is the key to help deliver real-time solutions for autonomous system. In particular, a faster computational platform can solve larger problems in the same time and solve the same problem more often in face of disruptions. The research on efficient stochastic control and optimization will focus on two aspects:

  • For the modeling side, we want to expand classes of problems for which one can generate approximation algorithms, which can be decomposed into massive subproblems to solve in parallel and efficiently.

  • For the computing side, we want to focus the work in designing practical framework based on the Fog Computing, which distributes computation, communication, control and storage closer to the end users. For instance, the distributed computing system based on connected mobile devices could perform massive parallel computing.

Large Scale Stochastic Optimization

This line of research roots in the fundamental mathematical modeling and algorithms designing based on convex optimization and stochastic programming, which aims to develop novel computational efficient algorithms to address uncertainty. This fundamental research plays an important role in the planning problem within large-scale real world application. Though a few stochastic modeling methods has been developed over the past several years, there is still a bottleneck in the design and realization of control laws for large-scale systems.

Fog Computing for Connected Large Scale Systems

Fog Computing is an advancement from the Cloud to take the benefit of the latests smart devices, which is powerful in computation, storage, sensing and communication. The major difference with the Cloud is that, Fog Computing is an architecture that distributes computation, communication, control and storage by using near-user edge devices, such as smart-phones, tablets, edge routers and connected UAVs, etc. This line of research focuses on the design of Fog framework for distributed control and parallel data analysis of large scale systems.

Fog framework takes the advantage of a collaborative multitude of end-user devices to perform parallel computing locally. Our algorithms identifies specific mathematical format, which enables us to decompose the original large-scale problem into massive subproblems. The decentralized optimization framework can automatically distribute the sub-problems to the available nodes in the network such that the computing efficiency is greatly improved through massive distributed computing.

UAVs Traffic Management (UTM)

As the UAV systems are actively integrated into our current national airspace system, the possibility of deployment of large groups of UAV closely cooperating together brings new potentialities for autonomous systems. The goal of UTM is to provide robust, reliable and efficient automating solutions. A major challenge of the UTM is to provide real-time robust solutions under uncertainty. The results are beneficial in numerous applications including cooperative surveillance, reconnaissance and monitoring tasks, search and rescue missions, searching for sources of pollution and sensory data acquisition. The research will focus on optimal control and network optimization for UAV trajectory planning, scheduling and formation control given the dynamics, communication, airspace capacity and operational constraints.

Learning-Based Control for Autonomous Systems

This research focuses on combining the optimization and modeling techniques from the artificial intelligence, such as deep learning and reinforcement learning, to develop algorithms for the large-scale optimal control problem. In practice, many autonomous systems are working based on data-driven models. Therefore, the learning-based approach with data-driven model could be a perfect combination. This exciting topic is rooted in the emergence of new opportunities for the Internet of Things and embedded artificial intelligence.